Sep 2023 – Apr 2025 · 1 yr 7 mos · Bengaluru, Karnataka, India

• Below are the highlights of my contributions:
• Frame Removal (Real-Time AI)
• Built a model to remove glasses from a user’s face in real-time—ideal for those wearing prescription glasses during virtual try-ons. Using MediaPipe for segmentation and a GAN for inpainting, I optimized this compute-heavy process with model pruning and GPU acceleration, hitting 20+ FPS on iPhones. I also developed a data pipeline for segmentation/inpainting datasets and compiled the model into CoreML, ONNX, and TF.js for cross-platform use.
• AR Glasses Web Try-On
• Created Lenskart’s first web-based AR try-on, achieving 20+ FPS on low-end devices and 60 FPS on high-end ones. Leveraging WebAssembly, TF.js, and GPU shaders, I ensured smooth 3D rendering with minimal latency.
• Contact Lens Try-On (Android, iOS, Web)
• Developed three SDKs—JavaScript (Web, TF.js), Java (Android, ONNX), and Swift (iOS, CoreML)—for Aqualens’ contact lens try-on. This multi-platform solution delivered real-time performance, slashing costs versus third-party alternatives.
• True Fit & Face Width Estimation
• Engineered True Fit to scale glasses to real-world dimensions using PyTorch regression models. Paired it with a TF.js-based face width estimator for real-time accuracy on the web.
• 3D Asset Improvement & Photoshoot AI
• Automated 3D asset fixes with Python scripts in Blender and built a PyTorch-based pipeline to generate photoshoot-grade images from raw glasses data.
• Glasses Segmentation & Store Lookr App
• Fine-tuned U-Net models for glasses segmentation (IoU/Dice metrics) and contributed to the 3D try-on app for Lenskart stores.
• Background Changing
• Used shaders and MediaPipe for live background switches during try-ons, maintaining high FPS.
• Tech Stack: TensorFlow, PyTorch, ONNX, CoreML, Swift, JavaScript, WebAssembly, Shaders, Python, MediaPipe, Java.

Sep 2022 – Sep 2023 · 1 yr · Bengaluru, Karnataka, India

May 2021 – May 2022 · 1 yr · Haldwani, Uttarakhand, India

• Lead Developer for Volunteer Management Applications: Spearheaded the development of comprehensive volunteer management platforms for Aam Aadmi Party and BJP in Uttarakhand using Expo React Native. These applications have been instrumental in enhancing volunteer coordination, communication, and engagement through intuitive mobile interfaces.
• Architect of Video and Image Processing System: Designed and implemented a high-performance video and image processing system tailored for volunteer activities. Utilizing AWS cloud services, FFmpeg for media manipulation, and Node.js for backend processing, this system supports efficient content creation and distribution workflows.
• Innovator in Social Media Automation: Developed an advanced automation framework using Puppeteer and Node.js to streamline content distribution across major social platforms including WhatsApp and Facebook. This solution significantly increased reach and engagement by automating routine publishing tasks.
• Creator of Global Dakiya Application: Engineered a cutting-edge news portal, Global Dakiya, with a focus on delivering personalized news content to users worldwide. Leveraged React JS for the front end to ensure a responsive, user-friendly interface, and utilized MongoDB and PostgreSQL for robust data management and storage.
• Technology Expertise:
• Frontend: React JS, React Native for cross-platform mobile applications.
• Backend: Node.js for scalable server-side solutions; extensive use of AWS for cloud-based infrastructure and services.
• Database Management: Proficient in MongoDB and PostgreSQL for comprehensive data storage, retrieval, and manipulation.
• Automation & Media Processing: Advanced experience with Puppeteer for web automation and FFmpeg for media processing tasks.
• Modern Web Technologies: Expertise in Rust and WebAssembly (WASM) for high-performance web applications.